使用OpenCV对二维码/条形码进行矫正

void correctImage(unsigned char *nv21, int width, int height, unsigned char *dest) {

    int resizeWidth = width/2;
    int resizeHeight = height/2;

    //初始化需要使用的Mat
    Mat imgMat(height * 3 / 2, width, CV_8UC1, nv21);
    //灰度图Mat
    Mat grayMat(height, width, CV_8UC1);
    //最终结果的Mat
    Mat resultMat(height, width, CV_8UC1,dest);
    //过程Mat
    Mat innerMat(resizeHeight, resizeWidth, CV_8UC1);
    Mat binMat(resizeHeight, resizeWidth, CV_8UC1);

    //将yuv420sp转为灰度图
    cvtColor(imgMat, grayMat, CV_YUV420sp2GRAY);

    //缩小图像尺寸以减少后面图像处理工作的耗时
    resize(grayMat, innerMat, Size(resizeWidth, resizeHeight));

    //自适应二值化参数
    int blockSize = 3;
	int constValue = 5;
	const int maxVal = 255;

    //自适应二值化
    adaptiveThreshold(innerMat, binMat, maxVal,ADAPTIVE_THRESH_GAUSSIAN_C, THRESH_BINARY,
		blockSize, constValue);

    int rectWidth = resizeWidth / 5;
    int rectHeight = resizeHeight / 5;
    int centerX = resizeWidth / 2, centerY = resizeHeight / 2;

    Mat element = getStructuringElement(2, Size(7, 7));

    //膨胀腐蚀
    for (int i = 0; i < 2; i++) {
        erode(innerMat, innerMat, element);
        i++;
    }

    erode(innerMat, binMat, element);
    binMat = innerMat - binMat;

    //寻找直线
    vector<Vec2f> lines;
    HoughLines(binMat, lines, 1, CV_PI / 150, 90, 0, 0);
    LOGV("OpenCV_CorrectImage寻找边线，边线数量:%d", lines.size());

    //由于边线可能有很多，这里最多取8条边线来计算角度
    int lineNum = lines.size() > 8 ? 8 : lines.size();
    float sum = 0;

    //num参数记录倾斜的线条的数量
    int num = 0;
    for (int i = 0; i < lineNum; i++) {
     
        float theta = lines[i][1];
        //各个角度对应的theta值,相当于:角度/180*PI
        //90->1.5708
        //100 -> 1.7453
        //80 -> 1.3963
        //95->1.6580
        //85->1.4835
        //0 ->0
        //5->0.0872
        //10 -> 0.1745
        //12  ->0.209
        //15 ->0.2618
        //180-> 3.1415
        //170 ->2.967

        // 过滤掉角度正常的，0-15度，80-100 ,170-180
        if ((theta >= 0 && theta <= 0.2618) || (theta >= 1.3963 && theta <= 1.7453) ||
            (theta >= 2.967 && theta <= 3.1416)) {
        } else {
            if (theta >= 1.5708)
                theta = theta - 1.5708;
            LOGV("OpenCV_CorrectImage寻找边线，角度:%f", theta);
            sum += theta;
            num += 1;
        }
    }

    //需要矫正
    if (num > 0) {
        float average = sum / num;
        //转成角度
        double angle = average / CV_PI * 180;
        LOGV("OpenCV_CorrectImage寻找边线，角度:%f", angle);

        //如果矫正角度大于5才矫正
        if (angle > 5) {
            Point2f center;
            center.x = float(width / 2.0);
            center.y = float(height / 2.0);
            Mat M = getRotationMatrix2D(center, angle, 1);
            //仿射变换，背景色填充为黑色
            warpAffine(grayMat, resultMat, M, resultMat.size(), 1, 0, Scalar(255));          
            LOGV("OpenCV_CorrectImage角度变换完成");
        }
    }
}